anhnh2002 / XTTSv2-Finetuning-for-New-Languages

60 stars 17 forks source link

Training new language sk=slovak language #17

Closed JohnF51 closed 2 weeks ago

JohnF51 commented 2 weeks ago

Hi Nguyễn Hoàng Anh I would like to ask you if you could help me and create a detailed tutorial for training a new language that is not yet supported by XTTSv2. The Slovak language is very similar to the Czech language which is already supported can this be used somehow? What would you recommend? I have Slovak voice wav data. How to modify your script so that I can use the Czech pre-trained model and modify it to the new "Slovak" language, Is it possible? Thank you for your great work.

anhnh2002 commented 2 weeks ago

Hi Nguyễn Hoàng Anh I would like to ask you if you could help me and create a detailed tutorial for training a new language that is not yet supported by XTTSv2. The Slovak language is very similar to the Czech language which is already supported can this be used somehow? What would you recommend? I have Slovak voice wav data. How to modify your script so that I can use the Czech pre-trained model and modify it to the new "Slovak" language, Is it possible? Thank you for your great work.

I think in your case there is no need to extend the tokenizer vocabulary. Just set the language to Czech by using --language="cs"